EDITORIAL Annie Darwin's Death, The Evolution Of Tuberculosis And The .

1428212223242526INTERNATIONAL JOURNAL OF EPIDEMIOLOGYEmergence of Mycobacterium tuberculosis with extensiveresistance to second-line drugs–worldwide, 2000–2004MMWR Morb Mortal Wkly Rep 2006;55:301–5.Cowley D, Govender D, February B et al. Recent and rapidemergence of W-Beijing strains of Mycobacterium tuberculosis in Cape Town, South Africa. Clin Infect Dis 2008;47:1252–59.Cooke GS, Hill AV. Genetics of susceptibility to humaninfectious disease. Nat Rev Genet 2001;2:967–77.Fellay J, Shianna KV, Ge D et al. A whole-genomeassociation study of major determinants for host controlof HIV-1. Science 2007;317:944–47.Telenti A. Adaptation, co-evolution, and human susceptibility to HIV-1 infection. Infect Genet Evol 2005;5:327–34.de Jong BC, Hill PC, Aiken A et al. Progression toactive tuberculosis, but not transmission, varies by27282930Mycobacterium tuberculosis lineage in the Gambia. J InfectDis 2008.Caws M, Thwaites G, Dunstan S et al. The influence ofhost and bacterial genotype on the development ofdisseminated disease with Mycobacterium tuberculosis.PLoS Pathog 2008;4:e1000034.Thwaites G, Caws M, Chau TT et al. Relationship betweenMycobacterium tuberculosis genotype and the clinical phenotype of pulmonary and meningeal tuberculosis. J ClinMicrobiol 2008;46:1363–68.Comas I, Gagneux S. The past and future of tuberculosisresearch. PLoS Pathog 2009;5:e1000600.Shanks N, Pyles RA. Evolution and medicine: the longreach of ‘‘Dr. Darwin’’. Philos Ethics Humanit Med 2007;2:4.

Published by Oxford University Press on behalf of the International Epidemiological Association! The Author 2009; all rights reserved. Advance Access publication 19 November 2009International Journal of Epidemiology 2009;38:1429–1439doi:10.1093/ije/dyp335REPRINTS AND REFLECTIONSMarriages between first cousins in Englandand their effects1George H DarwinI. The Proportion of First-CousinMarriages to all MarriagesDarwin GH, Marriages between first cousins in England andtheir effects. Fortnightly Review 1875; 24:22–41.1429Downloaded from http://ije.oxfordjournals.org/ by guest on February 1, 2015It is well known that when the Census Act, 1871, waspassing through the House of Commons, an attemptwas made by Sir J. Lubbock, Dr Playfair, and others,to have a question inserted with respect to the prevalence of cousin marriages, under the idea that whenwe were in possession of such statistics we should beable to arrive at a satisfactory conclusion as towhether these marriages are, as has been suspected,deleterious to the bodily and mental constitution ofthe offspring. It is unfortunately equally well knownthat the proposal was rejected, amidst the scornfullaughter of the House, on the ground that the idlecuriosity of philosophers was not to be satisfied.It was urged, that when we had these statistics itwould be possible to discover, by inquiry in asylums,whether the percentage of the offspring of consanguineous marriages amongst the diseased was greaterthan that in the healthy population, and thus tosettle the question as to the injuriousness of suchmarriages. The difficulty of this subsequent part ofthe inquiry was, I fear, much underrated by thosewho advocated the introduction of these questionsinto the census. It may possibly have been right toreject the proposal on the ground that every additional question diminishes the trustworthiness ofthe answers to the rest, but in any case the tonetaken by many members of the House shows howlittle they are permeated with the idea of the importance of inheritance to the human race.In the summer of 1873 the idea occurred to me thatit might be in some measure possible to fill up thishiatus in our national statistics. In looking throughthe marriages announced in the Pall Mall Gazette, Inoticed one between persons of the same surname;now as the number of surnames in England is verylarge, it occurred to me that the number of such marriages would afford a clue to the number of firstcousin marriages.In order to estimate what proportion of such marriages should be attributed to mere chance, I obtainedthe ‘‘Registrar-General’s Annual Report’’ for 1853,where the frequency of the various surnames isgiven. I here found that there were nearly 33,000surnames registered, and that the fifty commonestnames embraced 18 per cent. of all the population.It appear that one in 73 is a Smith, one in 76 aJones, one in 115 a Williams, one in 148 a Taylor,one in 162 a Davies, one in 174 a Brown, and thelast in the list is one Griffiths in 529. Now it isclear that in one marriage in 73 one of the partieswill be a Smith, and if there were no cause whichtended to make persons of the same surnamemarry, there would be one in 732, or 5,329 marriages,in which both parties were Smiths. Therefore theprobability of a Smith—Smith marriage due to mere1; similarly the chance of a Jones-Jones,chance is 5329a Davies-Davies and a Griffiths-Griffiths marriage11and 529would be 7612 , 16222 , respectively. And thesum of fifty such fractions would give the probabilityof a chance marriage, between persons of the samesurname, who owned one of these fifty commonestnames. The sum of these fifty fractions I find to be0.0009207, or 0.9207 per thousand. It might, however,be urged that if we were to take more than fifty of thecommon names, this proportion would be found to bemuch increased. I therefore drew a horizontal straightline, and at equal distances along it I erected ordi1The uppernates proportional to 7312 , 7612 , . . . , 5292,ends of these ordinates were found to lie in a curveof great regularity, remarkably like a rectangularhyperbola, of which my horizontal straight line wasone asymptote; and the ordinate corresponding toGriffiths was exceedingly short. Observing the greatregularity of the curve, I continued it beyond the fiftieth surname by eye, until it sensibly coincided withthe asymptote, at a point about where the hundredand twenty-fifth name would have stood, and then Icut out the whole (drawn on thick paper) andweighed the part corresponding to the fifty surnames,and the conjectural part. The conjectural addition wasfound to weigh rather more than one-tenth of theother part; and as the chance of same-name marriages is proportional to the areas cut out, I think Imay venture confidently to assert that in England andWales about one marriage in a thousand takes placein which the parties are of the same surname, andhave been uninfluenced by any relationship betweenthem bringing them together. Now it will appear

1430INTERNATIONAL JOURNAL OF EPIDEMIOLOGY(1) What proportion of this 1.25 per cent. weremarriages between first cousins.(2) What proportion marriages between first cousinsof the same surname bear to those between firstcousins of different surnames.If these two points could be discovered, the percentage of first-cousin marriages in the upper classes couldbe at once determined. I have endeavoured to find outthese proportions in several ways.An assistant was employed to count the marriagesof the men in the pedigrees of the English and Irishfamilies occupying about 700 pages of ‘‘Burke’sLanded Gentry,’’ marking every case where the marriage was ‘‘same-name’’. I then tried in every suchcase to discover, from a consideration of the pedigree,whether the marriage had been between first cousins.I found that in a certain number of cases I wasunable to discover this. The total number of pedigreesin the 700 pages was about 1,300; and of these Ihad to exclude 71, thinking that by only includingfamily trees where I could discover the relationshipof the parties, I should not obtain an unfair selectionof the whole. The marriages of the men alone wereincluded, because, had I included those of thewomen, many marriages would have been countedtwice over – once in the pedigree under consideration,and again in that of the husband. In this way, then,I found out of 9,549 marriages given by Burke 72were same-name first-cousin marriages, and 72 weresame-name marriages not between first cousins. Thisgives the percentage of same-name marriages as 1.5(not strikingly different from the 1.25 deduced fromthe Pall Mall Gazette), and of this percentage 0.75 is tobe attributed to first-cousin marriages.I further collected in the same way 1,989 marriagesfrom the ‘‘English and Irish Peerage,’’ and of these 18were same-name first-cousin marriages, or 0.9 percent. The number of same-name marriages notbeing first-cousin marriages was not, however, compared in this case. It will be observed, that theproportion is nearly 0.2 per cent. higher than withthe ‘‘Landed Gentry,’’ and as the nobility are knownto marry much inter se, this was perhaps to beexpected; however, 2,000 is too small a number onwhich to base a conclusion on this head with safety.The Peerage and Burke combined give 90 out of11,538, or 0.78 per cent., of same-name first-cousinmarriages.The next step was to send out a large number ofcirculars (about 800) to members of the upper middleand upper classes, in which I requested each personto give me the names of any members of the following classes, who married their first-cousins; viz.,(1) the uncles, aunts, father, and mother of theperson; (2) the brothers, sisters, and the person himself; (3) the first cousins of the person. I further askedfor the names of any persons in the above classeswho contracted same-name marriages not with firstcousins. I confined my question to near relations,because, had the more distant ones been included, arisk was run of getting a selected set of marriages – arisk which I am inclined to suspect was not avoided,as will hereafter appear.In about 300 of the circulars, I further asked for thetotal number of marriages contracted by the personsincluded in the Classes 1, 2, and 3. Care was taken toexclude, as far as possible, those persons who hadcousins in common, so that each answer shouldembrace a fresh field. I must here return my thanksto the many persons who so kindly filled in andreturned the circulars.The following result was obtained:TABLE st-CousinMarriages182Same name notFirst-CousinMarriages29From 181 circulars returned in which the totalnumber of marriages in each class was given, thefollowing was the result:TABLE BPercentage ofSame-Namemarriages,TotalTotalwhetherNumber of Percentage ofNumberCousin orFirst-Cousin First-Cousinofnot CousinMarriagesMarriages Marriages3,6631253.411.382Persons having no cousin marriages to fill in wereasked to return the circular blank in those caseswhere the total number of marriages was not askedfor. Of such blank returns, together with those whereDownloaded from http://ije.oxfordjournals.org/ by guest on February 1, 2015presently that far more than one marriage in a thousand is between persons of the same surname; and asI do not profess to have attained results of an accuracy comparable to 0.1 per cent., I am entitled to saythat same-name marriages, when they take place, aredue to consanguinity of the parties. If it permittedsuch accuracy, the method pursued would, however,include a compensation for this disturbing cause.With the help of an assistant the marriagesannounced in the Pall Mall Gazette in the years1869–72, and part of 1873, were counted, and werefound to be 18,528. Out of these 232 were betweenpersons of the same surname, that is 1.25 per cent.were same-name marriages. The same marriage isoccasionally announced twice over, but as there canbe no reason to suppose that this course has beenpursued oftener or seldomer with same-name marriages than with others, the result will not be vitiatedthereby. In order to utilise this result it now becamenecessary to determine-

MARRIAGES BETWEEN FIRST COUSINS IN ENGLAND AND THEIR EFFECTSSame ! name cousin ! marriages 142¼¼ :57All same ! name marriages249And in default of anything more satisfactory I amcompelled to accept this result as the first of my tworequisite factors.As to the second factor, - the proportion 2g: 1 fordifferent-name cousin marriages to same-namecousin marriages is, I fear, also unsatisfactory. Butbefore entering on this point I will indicate thesources of error in my returns:(1) The sensitiveness of persons in answering thequestion in cases where there are cousin marriages, particularly when any ill results mayhave accrued.(2) The non-return by persons who had no suchmarriages to fill in, and who would say, ‘‘Ihave no information, what is the use of returning this?’’3(3) The ignorance of persons of the marriages oftheir relations. This ignorance would be morelikely to affect the returns of different-namemarriages than of same-name ones. I feel convinced that this has operated to some extent, aswill be seen hereafter.(4) In the cases of same-name marriages, personswould be more likely to know of the marriagesbetween first cousins than of other such marriages. The discrepancy between Burke and mycirculars leads me to believe that this too hasoperated.I have been much surprised to find how very littlepeople know of the marriages of their relations,even so close as those comprised in my three classes.As it is clear that the marriages contracted by a man’suncles and aunts, and by his brothers and sisters,would be less likely to escape his notice than wouldthose contracted by his first cousins, I made an analysis of my circulars, including only the first twoclasses, viz: (1) uncles, aunts, father, and mother;(2) brothers and sisters and the person himself. Andthe results from this analysis made a nearer approachto those derived from Burke. But even then it seemedso unsatisfactory, that I feel sure that the indirectmethod, to which I now proceed, is on the wholemore reliable.It is possible to discover the proportion between thesame-name and different-name marriages in anentirely different way, and this I have tried to do.A man’s first cousins may be divided into fourgroups, viz: the children of (a) his father’s brothers,(b) of his father’s sisters, (c) his mother’s brothers,(d) his mother’s sisters. Of these four groups only(a) will in general bear the same surname as theperson himself. On the average the number of marriageable daughters in each family of each of the fourgroups will be the same. Were the four groups thenequally numerous, we might expect that the samename would bear to the different-name marriagesthe proportion of one to three. Since, however, aman cannot marry his sisters, this cannot holdgood; for the classes (a) and (d) are clearly on theaverage smaller than (b) and (c), and the proportionðaÞwe wish to discover is ðbÞþðcÞþðdÞ, which must evidently1be less than 3. To take a numerical example: A’sfather is one of 3 brothers, who married and havechildren, and A’s father had 2 sisters, who marriedand have children. A’s mother had 1 brother,who married and has children, and was one of 5sisters, who married and have children. Then clearlythe class(a) consists of 2 families.(b) consists of 2 families.(c) consists of 1 family.(d) consists of 4 families22So that the above fraction becomes 2þ1þ4 ¼ 7. Inthis case we may conclude that if A marries a firstcousin, it is 5 to 2 that he will marry one of a different surname. In another case the numbers mighthave been different, and therefore the fraction andthe betting also different. And what we wish to discover is the average value of this fraction. But for thevarious members of a large community there will be avery large number of such fractions, and some willoccur more frequently than others; so that in findingthis average value, each fraction should have itsproper weight assigned to it.In order to assign the weight to - say the abovefraction 27, we must take a thousand families andfind in how many of them there were 3 sons and 2daughters who married and had children, and in howmany there were 1 son and 5 daughters who marriedand had children. Having sufficiently indicated howthe required proportion depends on probabilities, Imay state that I sent out a number of circulars toDownloaded from http://ije.oxfordjournals.org/ by guest on February 1, 2015the total number of marriages was not given, 207 cameback to me; and the results derived from themwere found to agree closely with those in Table B.From Table A it is seen that there were 182 different-name cousin marriages to 66 same-name cousinmarriages; i.e. for every same-name cousin marriagethere were 2g different-name cousin marriages.And again there were 66 same-name cousin marriages to 29 same-name-not-cousin marriages; thatis rather more than two to one. This last result disagrees so much with that obtained from Burke andthe Peerage, where the proportion was, as abovestated, found to be as 1 to 1, that I am inclined tosuspect that I had either a run of luck against me, ormore probably that a considerable number of marriages between persons of the same surname, notbeing first cousins, escaped the notice of my correspondents. This latter belief is somewhat confirmedby what follows. If, however, I combine the resultsobtained from Burke with those from my circulars,I obtain the following:-1431

1432INTERNATIONAL JOURNAL OF EPIDEMIOLOGYmembers of the upper middle, and upper classes, andobtained and classified statistics with respect to aconsiderable number of families. I treated the question in four different ways. It might be supposed thata man, who had five families of first cousins in relation to himself, would be five times as likely to marrya first cousin as a man who had only one such family,or again it might be supposed that he would be onlyequally likely. The truth, however, will certainly liebetween these suppositions. The question, when treated from this point of view, leads to the result thatSame!name cousin!marriages1Different!name first!cousin marriages is greater than 4:44 andless thanabout 411.14:12.So that the true proportion would be41, so that the proportion would be reallyless than 4:141about 1; a result which differs but very slightly from46that given by the two other methods.The amount of arithmetical labour was so great thatI was obliged to make an approximation, whichwould, however, hardly affect the results, but as faras it went it would make the above fractions toosmall.I think on the whole it may be asserted, that thesame-name first-cousin marriages are to the differentname first-cousin marriages as 1 to 4. It ma

Mycobacterium tuberculosis HIV epidemic Urbanisation Globalisation Evolution Strain variation Drug resistance Human genetics Immunology Socio-economic conditions Housing TB control measures Evolution Co-evolution Figure 1 A 'systems epidemiology' approach to tuberculosis, which integrates demography, ecology and systems biology.